The negative punishment I use when clicker training my horses is withdrawal of access to the food treat when they mug. However, because the mugging involves a number of different behaviours, such as stretching neck, mouthing, pushing with nose, it can become difficult for them to discriminate which behaviour is resulting in the negative punishment consequences. So I have found training a behaviour that is incompatible with the unwanted behaviour is more effective and more resistant to extinction. No discrimination issues so less confusion and less trialling of random responses I don't appreciate. I train them to turn their head away to receive a treat and I practice it alot so the habit consolidates and the response is automatic.
The comments about conditioned responses in horses are very apposite. Horses are extremely adept and making associations between apparently random cues, responses and outcomes which they either value or want to avoid. Clever Hans is a good examplar of this. There is very interesting research on the role of dopamine in habit formation and it appears that well learned habits, which have become automatic are associated with the release of certain types of dopamine from the basal ganglia. And these responses can be reflexive (like salivation) or learned such as stopping in response to a classically conditioned seat cue. Animals trained to a higher number of correct avoidance (avoiding shock) responses show lower levels of behavioural and physiological fear than those trained to lower numbers. Wonder whether dopamine might be part of the reason?
As far as I understand it, in the aversive conditioning literature esp in rodents, responses are more rapidly acquired when the reinforcement is immediate or only briefly delayed after the performance of instrumental response. Given when we ride we mostly use aversive cues (pressure) we have an ethical obligation to minimise the intensity, frequency and duration of those cues by ensuring we practice immediacy in our reinforcement as much as possible. It would be interesting to be able to measure the duration of a horse's ability to remember a specific cue. It would appear that as grass eaters who run away from danger evolution has not exerted selection pressures for a long attention span. It would be also interesting to know whether horses perform negatively reinforced goal directed behaviour (action-outcome) as well as habitual behaviour (stimulus-response). Great thread!
Horses are blameless participants in training- its up to us to get it right